From October 1977 to December 1983, estrogen receptor (ER) and progesterone receptor (PR) levels were measured in 645 tumors from women with primary, unilateral, nonmetastatic breast cancer. All of them were treated surgically. Some received adjuvant radiotherapy, adjuvant chemotherapy, or adjuvant hormonotherapy. A level of greater than 5 fmol/mg cytosolic protein was considered as positive for both ER and PR. Unifactorial analysis, using Kaplan and Meier estimates and the log-rank test, revealed that disease-free survival (DFS) and overall survival (SV) were both strongly related to age, tumor size, nodal status, nodal effraction, histopathologic grading (SBR), ER, and PR. Menopausal status and number of intramammary tumor foci were not significant. Multifactorial analysis, using Cox's model, revealed a strong relationship between SV and age (poor prognosis [pp]: less than or equal to 37 years old), menopausal status (pp: postmenopausal) tumor size, nodal status (pp: N+ greater than 3), nodal effraction, ER (pp: less than or equal to 5 fmol/mg), histopathologic grading (pp: SBR = 3), and PR (pp: less than or equal to 5 fmol/mg). Similarly, multifactorial analysis revealed a strong correlation between DFS and age, tumor size, nodal status, nodal effraction, histopathologic grading, and PR. A prognostic score taking into account these prognostic factors was calculated for DFS and SV. Analysis of this score allowed us to divide our patients into four significantly different (P less than 0.0001) groups with high, intermediate, and low risk of relapse. Our procedure was then validated using the sample test technique. These results show that both ER and PR have their own prognostic weight and should be considered, among other classic prognostic factors, when adjuvant therapies are indicated after surgical treatment of breast cancer.